SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMs

• URL: http://arxiv.org/abs/2506.04250v1
• Date: Sun, 01 Jun 2025 01:19:37 GMT
• Title: SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMs
• Authors: Shaona Ghosh, Amrita Bhattacharjee, Yftah Ziser, Christopher Parisien,
• Abstract summary: This paper investigates an approach called SafeSteer for guiding the outputs of large language models (LLMs)<n>We employ a simple, gradient-free unsupervised method to enhance safety steering while preserving text quality, topic relevance, and without explicit refusal.
• Score: 7.120986296945107
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Fine-tuning large language models (LLMs) to adapt to evolving safety policies is costly and impractical. Mechanistic interpretability enables inference-time control through latent activation steering, yet its potential for precise, customizable safety adjustments remains largely untapped. This paper investigates an approach called SafeSteer for guiding the outputs of LLMs by: (i) leveraging category-specific steering vectors for more precise control, (ii) employing a simple, gradient-free unsupervised method to enhance safety steering while preserving text quality, topic relevance, and without explicit refusal, and (iii) accomplishing this without a hard requirement of contrastive pairwise safe data. We also highlight that our method, being simple and effective, aligns with recent studies suggesting that simple techniques often outperform more complex ones in activation steering. We showcase the effectiveness of our approach across various LLMs, datasets, and risk categories, demonstrating its ability to provide precise control, prevent blanket refusals, and guide models toward generating safe content while maintaining topic relevance.

Related papers

• AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint [49.641959856967276]
  We present a theoretically grounded and empirically effective activation steering method called AlphaSteer.<n>For utility preservation, it learns to construct a nearly zero vector for steering benign data, with the null-space constraints.<n>Experiments across multiple jailbreak attacks and utility benchmarks demonstrate the effectiveness of AlphaSteer.
  arXiv  Detail & Related papers  (2025-06-08T07:03:28Z)
• Shape it Up! Restoring LLM Safety during Finetuning [66.46166656543761]
  Finetuning large language models (LLMs) enables user-specific customization but introduces critical safety risks.<n>We propose dynamic safety shaping (DSS), a framework that uses fine-grained safety signals to reinforce learning from safe segments of a response while suppressing unsafe content.<n>We present STAR-DSS, guided by STAR scores, that robustly mitigates finetuning risks and delivers substantial safety improvements across diverse threats, datasets, and model families.
  arXiv  Detail & Related papers  (2025-05-22T18:05:16Z)
• Graphormer-Guided Task Planning: Beyond Static Rules with LLM Safety Perception [4.424170214926035]
  We propose a risk-aware task planning framework that combines large language models with structured safety modeling.<n>Our approach constructs a dynamic-semantic safety graph, capturing spatial and contextual risk factors.<n>Unlike existing methods that rely on predefined safety constraints, our framework introduces a context-aware risk perception module.
  arXiv  Detail & Related papers  (2025-03-10T02:43:54Z)
• Vulnerability Mitigation for Safety-Aligned Language Models via Debiasing [12.986006070964772]
  Safety alignment is an essential research topic for real-world AI applications.<n>Our study first identified the difficulty of eliminating such vulnerabilities without sacrificing the model's helpfulness.<n>Our method could enhance the model's helpfulness while maintaining safety, thus improving the trade-off-front.
  arXiv  Detail & Related papers  (2025-02-04T09:31:54Z)
• Towards Inference-time Category-wise Safety Steering for Large Language Models [3.712541089289745]
  Large language models (LLMs) have seen unprecedented advancements in capabilities and applications across a variety of use-cases. The fragile nature of LLMs warrants additional safety steering steps via training-free, inference-time methods. Unlike recent inference-time safety steering works, in this paper we explore safety steering of LLM outputs using category-specific steering vectors.
  arXiv  Detail & Related papers  (2024-10-02T02:02:06Z)
• SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering [56.92068213969036]
  Safety alignment is indispensable for Large Language Models (LLMs) to defend threats from malicious instructions.<n>Recent researches reveal safety-aligned LLMs prone to reject benign queries due to the exaggerated safety issue.<n>We propose a Safety-Conscious Activation Steering (SCANS) method to mitigate the exaggerated safety concerns.
  arXiv  Detail & Related papers  (2024-08-21T10:01:34Z)
• Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training [67.30423823744506]
  We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position.<n>DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence.
  arXiv  Detail & Related papers  (2024-07-12T09:36:33Z)
• Towards Comprehensive Post Safety Alignment of Large Language Models via Safety Patching [74.62818936088065]
  textscSafePatching is a novel framework for comprehensive PSA.<n>textscSafePatching achieves a more comprehensive PSA than baseline methods.<n>textscSafePatching demonstrates its superiority in continual PSA scenarios.
  arXiv  Detail & Related papers  (2024-05-22T16:51:07Z)
• Closing the Closed-Loop Distribution Shift in Safe Imitation Learning [80.05727171757454]
  We treat safe optimization-based control strategies as experts in an imitation learning problem. We train a learned policy that can be cheaply evaluated at run-time and that provably satisfies the same safety guarantees as the expert.
  arXiv  Detail & Related papers  (2021-02-18T05:11:41Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.